We address the assignment of ICD-10 codes for causes of death by analyzing free-text descriptions in death certificates, together with the associated autopsy reports and clinical bulletins, from the Portuguese Ministry of Health. We leverage a deep neural network that combines word embeddings, recurrent units, and neural attention, for the generation of intermediate representations of the textual contents. The neural network also explores the hierarchical nature of the input data, by building representations from the sequences of words within individual fields, which are then combined according to the sequences of fields that compose the inputs. Moreover, we explore innovative mechanisms for initializing the weights of the final nodes of the network, leveraging co-occurrences between classes together with the hierarchical structure of ICD-10. Experimental results attest to the contribution of the different neural network components. Our best model achieves accuracy scores over 89%, 81%, and 76%, respectively for ICD-10 chapters, blocks, and full-codes. Through examples, we also show that our method can produce interpretable results, useful for public health surveillance.

Binary file ES_Abstracts_Final_ECDC.txt matches

No abstract

No abstract

BackgroundVaccination is the key measure available for prevention of the public health burden of annual influenza epidemics. This article describes national trends in seasonal influenza vaccine (IV) coverage in Portugal from 1998/99 to 2010/11, analyzes progress towards meeting WHO 2010 coverage goals, and addresses the effect of major public health threats of the last 12 years (SARS in 2003/04, influenza A (H5N1) in 2005/06, and the influenza A (H1N1)2009 pandemic) on vaccination trends.MethodsThe National Institute of Health surveyed (12 times) a random sample of Portuguese families. IV coverage was estimated and was adjusted for age distribution and country region. Independence of age and sex coverage distribution was tested using a modified F-statistic with a 5% significance level. The effect of SARS, A (H5N1), and the A (H1N1)2009 pandemic was tested using a meta-regression model. The model was adjusted for IV coverage in the general population and in the age groups.ResultsBetween 1998/99 and 2010/11 IV, coverage in the general population varied between 14.2% (CI 95%: 11.6%–16.8%) and 17.5% (CI 95%: 17.6%–21.6%). There was no trend in coverage (p = 0.097). In the younger age group (<15 years) a declining trend was identified until 2008/09 (p = 0.005). This trend reversed in 2009/10. There was also a gradual and significant increase in seasonal IV coverage in the elderly (p for trend < 0.001). After 2006/07, IV coverage remained near 50%. Adjusting for baseline trends, there was significantly higher coverage in the general population in 2003/04 (p = 0.032) and 2005/06 (p = 0.018). The high coverage observed in the <15-year age group in season 2009/10 was also significant (p = 0.015).ConclusionsIV coverage in the elderly population displayed an increasing trend, but the 75% WHO 2010 target was not met. This result indicates that influenza vaccination strategy should be improved to meet the ambitious WHO coverage goals. The major pandemic threats of the past decade had a modest but significant effect on seasonal influenza vaccination. There was an increase in vaccine uptake proportion in the general population in 2003/04 and in 2005/06, and in individuals <15 years old in 2009/10.